基于机器学习的绝热层材料粘-超弹性本构模型参数确定方法

Parameter determining method for visco-hyperelastic constitutive model of insulation material based on machine learning

为了更准确地描述固体发动机绝热层大变形特点以及力学响应与速率相关的力学性能,建立了一种粘-超弹性本构模型。模型由超弹性与粘弹性两部分组成,超弹部分的应力响应基于指数-对数形式的应变能密度函数推导获得,粘弹部分的应力响应为卷积积分形式。引入机器学习方法用于本构模型参数的厘定过程,该方法可以实现复杂模型的参数确定。将绝热层材料的试验数据与模型预测结果进行对比分析。结果表明,模型能够准确描述绝热层材料的力学响应,在不同拉速下(1.98、99、1980 mm/min)模型预测结果与试验结果的平均偏差分别为1.74%,8.1%,3.5%,从而说明基于机器学习的模型参数确定方法具有较高的可靠性与便捷性。

In order to more accurately describe the large deformation characteristics of solid rocket motor insulation and the rate-dependent mechanical properties,a visco-hyperelastic constitutive model was established.The model consists of hyperelastic part and viscoelastic parts.The stress response of hyperelastic part was derived based on the strain energy density function in exponential-logarithmic form.The stress response of viscoelastic part was expressed in convolution integral form.The machine learning method was applied to determine the parameters of constitutive model,and the parameters of complex model can be determined by this method.The experimental data of insulation material was compared and analyzed with the model prediction results.The results show that the model can accurately describe the mechanical response of insulation material.The average deviations between the predicted results and the test data under different tensile velocities(1.98,99,1980 mm/min)are 1.74%,8.1%and 3.5%,respectively,which indicates that the method of determining model parameters based on machine learning has high reliability and convenience.